[
http://opencast.jira.com/browse/MH-9225?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=32704#comment-32704
]
Tobias Wunden edited comment on MH-9225 at 1/8/13 10:35 PM:
------------------------------------------------------------
I can see a race condition where there is a preprocessing workflow going on
with the three "operations"
* scheduled
* capturing
* ingesting
Assuming that the capture agent is recording and the workflow has been
forwarded to "capturing", the capture agent starts ingesting (and during that
time for some reason is not updating the status). At the end of the ingest
operation, the ingest service adds the "real" workflow steps to that
preprocessing workflow (see IngestServiceImpl lines 743 ff) and forwards the
workflow to the first real workflow operation by calling
workflowinstance.next() until that operatio is reached.
Just while (or right before) the ingest service is extending its local copy of
the workflow with the real workflow steps, the capture agent finally gets
around to reporting the ingested status to the capture agent state service and
the agent state service is calling resume() on the workflow instance at
CaptureAgentStateImpl line 542. This will move the current workflow forward to
the "capturing" step, asking the workflow service to start processing that
operation. The operation is scheduled and eventually dispatched, returning
Action.PAUSE as the action.
was (Author: twunden):
I can see a race condition where there is a preprocessing workflow going on
with the three "operations"
* scheduled
* capturing
* ingesting
Assuming that the capture agent is waiting and the workflow has been
initialized with "scheduled", the capture agent is first recording, then
ingesting (and during that time for some reason not updating the status). At
the end of the ingest operation, the ingest service adds the "real" workflow
steps to that preprocessing workflow (see IngestServiceImpl lines 743 ff) and
forwards the workflow to the first real workflow operation by calling
workflowinstance.next() until that operatio is reached.
Just while (or right before) the ingest service is extending its local copy of
the workflow with the real workflow steps, the capture agent finally gets
around to reporting the ingested status to the capture agent state service and
the agent state service is calling resume() on the workflow instance at
CaptureAgentStateImpl line 542. This will move the current workflow forward to
the "capturing" step, asking the workflow service to start processing that
operation. The operation is scheduled and eventually dispatched, returning
Action.PAUSE as the action.
> Workflows automatically go into hold state after inspect but resume
> processing correctly with a nudge from the resume workflow endpoint
> ---------------------------------------------------------------------------------------------------------------------------------------
>
> Key: MH-9225
> URL: http://opencast.jira.com/browse/MH-9225
> Project: Matterhorn Project
> Issue Type: Bug
> Components: Administrative Tools
> Affects Versions: 1.4, 1.3, 1.2
> Reporter: Adam McKenzie
> Assignee: Adam McKenzie
> Priority: Critical
>
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
http://opencast.jira.com/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
_______________________________________________
Matterhorn mailing list
Matterhorn@opencastproject.org
http://lists.opencastproject.org/mailman/listinfo/matterhorn
To unsubscribe please email
matterhorn-unsubscr...@opencastproject.org
_______________________________________________
